Hydrological forecasting plays a vital role in drought management, water resources planning, hydropower operations, and flood control. Growing climatic variability has intensified the demand for forecasts that are both accurate and reliable. To this end, hydrological models depend on high-quality meteorological forecasts as inputs. Nevertheless, commonly used raw ensemble forecasts are often biased and exhibit insufficient spread, which makes statistical postprocessing necessary to improve practical usefulness. Member-by-member postprocessing (MBMP) is a widely used approach, but the standard linear corrections treat ensemble members with large deviations from the mean in the same way as those close to it, which is an assumption that may not always be appropriate. This study explores nonlinear alternatives to MBMP for temperature forecasts, motivated by the need to improve both predictive accuracy and practical utility in weather and hydrological applications. Our objective was to test whether nonlinear transformations could offer advantages over traditional linear corrections. The linear functions in MBMP were replaced with logarithmic, radical, and exponential transformations, and a novel angular formulation was designed to better handle member variability. These methods were applied to ensemble ECMWF forecasts for 32 basins across Québec, with lead times of 2, 5, and 9 days. The results demonstrate that nonlinear MBMP variants perform competitively with existing methods in terms of accuracy, and in particular those based on logarithmic and radical transformations, while the angular approach shows promise in enhancing forecast reliability, in addition to generally outperforming the standard MBMP at a 2-day lead time.